Abstract：The segmentation of vessels in lung CT images plays an important role in the diagnosis and surgical treatment of diseases. In medical image related tasks, deep learning is widely used due to their strong feature representation and discriminative learning abilities. However, deep learning-based methods requires expensive GPUs and large amount of labeled data. In order to achieve better balance between the accuracy and efficiency of vessel segmentation in lung CT images, in this paper, a faster and more effective unsupervised vessel segmentation algorithm based onmulti-dimensional information fusion (MDF) was proposed. The algorithm designed 2D segmentation branches and 3D segmentation branches to make full use of 2D and 3D information, and combined the results of multiple branches in the final segmentation results, which can be quickly and effectively integrated into the traditional unsupervised algorithm. Meanwhile, MDF has strong parallelism capability and can be significantly accelerated on GPU. The proposed MDF based vessel segmentation algorithm was evaluated on the challenging VESSEL12 dataset and CARVE14 dataset comprehensively. The experimental results showed that MDF segmented the vessels in lung CT images with higher accuracy compared with other unsupervised methods. On the CARVE14 dataset, the DSC coefficient of vessels reached 0.716. Moreover, the MDF inference speed was roughly 20 times faster than Frangi algorithms, a multiscale algorithm based on Hession matrix, by GPU parallel optimization. Compared with deep learning-based methods, the segmentation performance of MDF exhibited stronger adaptation ability.
[1] Verschakelen JA, Fraeyenhoven LV. Pulmonary vascular diseases[J]. Current Opinion in Radiology, 1990, 2(3):360-364. [2] Mourani PM, Sontag MK, Younoszai A, et al. Clinical utility of echocardiography for the diagnosis and management of pulmonary vascular disease in young children with chronic lung disease[J]. Pediatrics, 2008, 121(2):317-325. [3] Jones PL, Cowan KN, Rabinovitch M. Tenascin-C, proliferation and subendothelial fibronectin in progressive pulmonary vascular disease[J]. The American Journal of Pathology, 1997, 150(4): 1131-1342. [4] El-Baz A, Beache GM, Gimel G, et al. Computer-aided diagnosis systems for lung cancer: challenges and methodologies[J]. International Journal of Biomedical Imaging, 2013, 2013: 475-520. [5] Korfiatis P, Karahaliou A, Costaridou L. Automated vessel tree segmentation: challenges in computer aided quantification of diffuse parenchyma lung diseases[C]//2009 9th International Conference on Information Technology and Applications in Biomedicine. Boston: IEEE, 2009: 1-4. [6] Zhu X, Xue Z, Gao X, et al. Voles: Vascularity-oriented level set algorithm for pulmonary vessel segmentation in image guided intervention therapy[C]//2009 IEEE International Symposium on Biomedical Imaging: From Nano to Macro.Boston: IEEE, 2009: 1247-1250. [7] 于洋. 肺部CT血管分割及3D重建[D]. 哈尔滨:哈尔滨工业大学, 2016. [8] Frangi AF, Niessen WJ, Vincken KL, et al. Multiscale vessel enhancement filtering[C]// First International Conference on Medical Image Computing and Computer-Assisted Intervention. Cambridge: Springer Berlin Heidelberg, 1998: 130-137. [9] 张新红, 张帆, 崔延斌.基于多尺度自适应滤波的血管增强[J].计算机工程与应用, 2015, 51(14):179-185. [10] Jerman T, Perun F, Likar B, et al. Beyond Frangi: an improved multiscale vesselness filter[J]. Medical Imaging, 2015, 9413:623-633. [11] Sato Y, Westin Tissue classification based on 3D local intensity structures for volume rendering[J]. IEEE Transactions on Visualization & Computer Graphics, 2000, 6(2):160-180. [12] 高齐新, 杨金柱, 赵大哲, 等.一种基于算子的-肺部血管分割算法[J].系统仿真学报, 2008, 20(20):5534-5537. [13] Shikata H, McLennan G, Hoffman EA, et al. Segmentation of pulmonary vascular trees from thoracic 3D CT images[J]. International Journal of Biomedical Imaging, 2009, 1(1):1-11. [14] 王文明, 孙丰荣, 刘炜, 等.基于形态学与局部特征结构的肺血管树重建[J].计算机工程, 2009, 35(19):200-202. [15] Babu JS, Mole S. Automatic vessel segmentation of lung affected patterns in MDCT using decision tree classification[J]. Middle-East Journal of Scientific Research, 2014, 22(11):1679-1685. [16] Chen Bin. Automatic segmentation of pulmonary blood vessels and nodules based on local intensity structure analysis and surface propagation in 3D chest CT images[J]. International Journal of Computer Assisted Radiology and Surgery, 2011, 7(3):465-482. [17] Litjens G, Kooi T, Bejnordi BE, et al. A Survey on deep learning in medical image analysis[J]. Medical Image Analysis, 2017, 42(9):60-88. [18] Harrison AP, Xu Z, George K, et al. Progressive and multi-path holistically nested neural networks for pathological lung segmentation from CT images[C]// 20th International Conference on Medical Image Computing and Computer Assisted Intervention. Quebec City: Springer International Publishing, 2017: 621-629. [19] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[C]//18th International Conference Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015. Munich: Springer International Publishing, 2015: 234-241. [20] Çiçek Ö, Abdulkadir A, Lienkamp SS, et al. 3D U-Net: learning dense volumetric segmentation from sparse annotation[C]//Medical Image Computing and Computer-Assisted Intervention-MICCAI 2016: 19th International Conference. Athen: Springer International Publishing, 2016: 424-432. [21] Xiao X, Lian S, Luo Z, et al. Weighted res-unet for high-quality retina vessel segmentation[C]//2018 9th international Conference on Information Technology in Medicine and Education (ITME). Hangzhou: IEEE, 2018: 327-331. [22] Zhou Z, Rahman Siddiquee MM, Tajbakhsh N, et al. Unet++: a nested U-net architecture for medical image segmentation[C]//Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th International Workshop, DLMIA 2018, and 8th International Workshop, ML-CDS 2018. Granada: Springer International Publishing, 2018: 3-11. [23] Huang H, Lin L, Tong R, et al. Unet 3+: a full-scale connected Unet for medical image segmentation[C]//ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Barcelona: IEEE, 2020: 1055-1059. [24] Abdollahi A, Pradhan B, Alamri AM. VNet: an end-to-end fully convolutional neural network for road extraction from high-resolution remote sensing data[J]. IEEE Access, 2020, 8(8):179424-179436. [25] 孙若凡, 张唯唯. 基于3D U-Net实现人体耳软骨MRI图像的解剖结构分割[J].中国生物医学工程学报, 2021, 40(5):531-539. [26] 艾琦, 王军, 任福全, 等.基于尺寸自适应深度神经网络的胸部图像肺结节检测[J].中国生物医学工程学报, 2021, 40(6):10-14. [27] Jin Q, Chen Q, Meng Z, et al. Construction of retinal vessel segmentation models based on convolutional neural network[J]. Neural Processing Letters, 2020, 52(2):1005-1022. [28] Qin Y, Zheng H, Gu Y, et al. Learning tubule-sensitive cnns for pulmonary airway and artery-vein segmentationin CT[J]. IEEE Transactions on Medical Imaging, 2021, 40(6): 1603-1617. [29] Dai J, Qi H, Xiong Y, et al. Deformable convolutional networks[C]//Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 764-773. [30] Otsu N. A threshold selection method from gray-level histograms[J]. IEEE Transactions on Systems Man & Cybernetics, 2007, 9(1):62-66. [31] Filho P, Cortez PC, Barros A, et al. Novel and powerful 3D adaptive crisp active contour method applied in the segmentation of CT lung images[J]. Medical Image Analysis, 2017, 35(35):503-516. [32] Rudyanto RD, Kerkstra S, Van Rikxoort EM, et al. Comparing algorithms for automated vessel segmentation in computed tomography scans of the lung: the VESSEL12 study[J]. Medical Image Analysis, 2014, 18(7):1217-1232. [33] Charbonnier JP, Brink M, Ciompi F, et al. Automatic pulmonary artery-vein separation and classification in computed tomography using tree partitioning and peripheral vessel matching[J]. IEEE Transactions on Medical Imaging, 2015, 35(3):882-892. [34] Association W Declaration of Helsinki Ethical principles for medical research involving human subjects. [J]. Journal of the Indian Medical Association, 2009, 14(1):233-238. [35] Sato Y, Nakajima S, Shiraga N, et al. Three-dimensional multi-scale line filter for segmentation and visualization of curvilinear structures in medical images[J]. Medical Image Analysis, 1998, 2(2):143-168.
